The demo contained in this package is described in 

Balanced Exploration and Exploitation Model Search for Efficient Epipolar Geometry Estimation
Liran Goshen, Ilan Shimshoni
IEEE PAMI, Vol 30, No 7, 1230-1242, July 2008.

Please refer to the paper for details

The application can be run using the command line or using a matlab interface.
In the BEEM_Example directory are BEEM.m and BEEM2.m which can be used to run the program in a 
similar way as from the command line but in a more convinient manner.


To run in the command line use following syntax:

1. For two input images "Filename1.pgm" and "Filename2.pgm" use

BEEM 1 Filename1.pgm Filename2.pgm

or

BEEM 1 Filename1.pgm Filename2.pgm D TR K L iters

Those two options extract keypoints from images "Filename1.pgm" and "Filename2.pgm".
Match them by LSH, based on those keypoint matches, calculate F and supported inliers and save those to files F.txt & matches.txt respectively.



2. In the case when the correspondences were previously extracted and saved in "p1_m", "p2_m", "rr1", "rr2" and "distanceRatio" use

BEEM 2 1 TR iters D

This option receives putative matches "p1_m", "p2_m", their SIFT parameters "rr1", "rr2" and the distance ratio "distanceRatio" from files. 
rr1, rr2 -  SIFT parameters for the putative match. rr1 for p1_m and rr2 for p2_m. 
The p1_m/p2_m files contain the putative correspondences (x,y coordinates).
The first column in rr1/rr2 is the rotation angle of the dominant gradient of the SIFT descriptor.
The second column in rr1/rr2 is the scale in the space scale of the SIFT descriptor. 
The distanceRatio - the distance ratio of the closest to the second-closest keypoints.
It calculates F and supported inliers and saves those to files F.txt & matches.txt respectively.


3. In the case when the correspondences were previously extracted and saved in "p1_m", "p2_m", "rr1", "rr2", and their probabilities 
were calculated and saved in "probabilities", use:

BEEM 2 2

or

BEEM 2 2 TR iters

Those options are similar to the previous one, but receive from file probabilities and not the distance ratios. 



Optional parameters:
D - Maximal distance ratio between the first and the second guess (Default value = 0.8)
TR - Epipolar distance (Default value - 5)
K - LSH parameter (Default value - 12)
L - LSH parameter (Default value - 15)
iters - maximal number of iterations (Default value - 500)